In the monkey, stimulus memory, including stimulus recognition and stimulus-reward association, among others, has now been found to depend on interaction between the sensory processing streams (e.g. the occipitotemporal visual pathway) and a specific limbic-system circuit, viz. the rhinal circuit, consisting of the rhinal cortex, orbitofrontal cortex, and the magnocellular division of the thalamic medial dorsal nucleus. By contrast, the hippocampal circuit, including the cingulate gyrus, anterior thalamic nuclei, and mamillary bodies, appears to play little if any role in these particular memory processes; indeed, outside of spatial memory deficits, there is little evidence yet of any memory loss in monkeys after damage to the hippocampal circuit. Interestingly, however, patients with very early and selective hippocampal injury due to hypoxic/ischemic episodes show pronounced deficits in stimulus memory, but of a particular form termed 'episodic memory', i.e. autobiographical memory for recent events; at the same time, their 'semantic memory', i.e. their memory for facts or generic knowledge about the world, appears to be relatively spared. Comparison of their pattern of mnemonic deficits with those of monkeys with selective limbic lesions suggests that the rhinal and hippocampal circuits form a neural hierarchy, with the stimulus memory functions of the rhinal circuit being sufficient to support semantic but not episodic memory, which must therefore require, in addition, the still higher-order contributions of the hippocampal circuit. The anatomical evidence supports this notion, in that the cortical sensory processing streams and the hippocampus can communicate with each other only through a link provided by the rhinal cortex. Although rhinal cortical lesions nearly preclude monkeys from recognizing objects presented a minute or two earlier, they do not prevent very short-term memory; indeed, memory of a new object for five seconds or less is essentially normal after rhinal lesions. This residual ability can be demonstrated by training monkeys with rhinal lesions to perform delayed non-matching-to-sample (DNMS) with trial-unique stimuli and short intratrial delays. Under these conditions the lesioned monkeys appear to learn DNMS not as a cognitive principle based on recognition memory but as a behavioral rule based on habit formation. This type of rule learning in the absence of the rhinal cortex depends on interaction between the occipitotemporal visual processing stream and the ventral frontal cortex, the ability being abolished when lesions of this cortex (including inferior frontal and orbitofrontal) are superimposed on rhinal lesions. A simpler type of habit formation, namely, concurrent visual discrimination learning with 24-hour intertrial intervals, depends instead on interaction between the cortical visual processing stream and its neostriatal targets in the ventrocaudal putamen and tail of caudate nucleus; our new evidence indicates that impairment in this form of learning is directly proportional to the extent of damage to these targets in the neostriatum.